中文摘要：

现有基于小波变换或对比度的能见度检测算法，计算准确性容易受到噪声影响，且时效性差。为此，该文提出一种基于路面视亮度差平方最优化的视频能见度检测算法。首先由摄像机标定参数，提取出亮度和高度一致的路面兴趣域，进而求得路面的视亮度作为真实参考值；根据 Koschmieder 理论，选定消光系数的初值，将由公式得到的视亮度值与真实参考值作差平方，建立目标函数；最后利用最小二乘法逼近原理，求目标函数取得最小值时消光系数的最优解。经过实验论证，该算法具有计算精度高、速度快、对噪声鲁棒性好等优点，应用前景广泛。

英文摘要：

In order to solve the deficiency of existing visibility detection methods, this paper presents a novel algorithm by working out the least square calculated from apparent luminance of roads. First, on the basis of camera self-calibration, the region of interest with constant height and even illumination is extracted. Then apparent luminance values of roads are figured out as real reference values. After that, an initial value of extinction coefficient is selected properly, with which a series of apparent luminance values are calculated according to Koschmieder theory. Finally, the objective function of extinction coefficient optimization is established based on the least square between calculated values and reference values of apparent luminance. Experimental results show that the proposed visibility detection algorithm has high accuracy, fast speed, fine robustness and broad application prospects.